Macs
Which AI models run on the M5 Ultra (512 GB, 80-core GPU)?
The M5 Ultra (512 GB, 80-core GPU) has 512 GB of memory, about 384 GB of it usable for a model. 109 of the 110 models in our catalog run on it at 4-bit or better.
Data as of 2 November 2026 · Vendor specs
512 GBmemory
~384 GB usable for a model
—bandwidth: not published
109models run at 4-bit or better
Models that run on it
| Model | Best build | Memory | Speed |
|---|---|---|---|
| GLM-5.3-Flash320B MoE | Q4_K_M | 200 GB | no estimate |
| DeepSeek-V4-Flash284B MoE | Q4_K_M | 177 GB | no estimate |
| Qwen3-235B-A22B235B MoE | Q4_K_M | 134 GB | no estimate |
| Qwen3.8-Flash-Next125B MoE | Q4_K_XL | 115 GB | ~111 tok/s ModelFit est. |
| Nemotron 3 Super 120B-A12B120B MoE | Q4_K_M | 90 GB | ~61 tok/s ModelFit est. |
| Nemotron 3.5 Super VL 120B-A12B120B MoE | Q4_K_M | 76 GB | no estimate |
| gpt-oss-120b117B MoE | MXFP4 | 68 GB | ~146 tok/s ModelFit est. |
| Llama 3.3 70B70.6B | Q8_0 | 78 GB | ~13 tok/s ModelFit est. |
| Llama 3.3 Nemotron Super 49B49B | Q4_K_M | 31 GB | no estimate |
| Qwen3.6-35B-A3B35B MoE | Q8_0 | 41 GB | ~160 tok/s ModelFit est. |
| Gemma 4 31B30.7B | Q8_0 | 33 GB | ~28 tok/s ModelFit est. |
| Qwen3-Coder-30B-A3B30.5B MoE | Q8_0 | 34 GB | ~130 tok/s ModelFit est. |
| Nemotron 3 Nano 30B-A3B30B MoE | Q4_K_M | 19.6 GB | no estimate |
| Qwen3.8-27B27.3B | Q8_0 | 29 GB | ~34 tok/s ModelFit est. |
| Gemma 4 26B-A4B25.2B MoE | Q8_0 | 30 GB | ~123 tok/s ModelFit est. |
| gpt-oss-20b21B MoE | MXFP4 | 15.1 GB | ~167 tok/s ModelFit est. |
| Nemotron Nano 9B v29B | Q4_K_M | 6.6 GB | no estimate |
| Qwen3.5 9B Instruct9B | Q8_0 | 12.0 GB | ~96 tok/s ModelFit est. |
| Qwen2.5 7B Instruct7B | Q4_K_M | 6.6 GB | ~153 tok/s ModelFit est. |
| Gemma 3 4B Instruct4B | Q8_0 | 5.0 GB | ~159 tok/s ModelFit est. |
| Ling-3.1-flash-1019560B | Q4_K_M | 368 GB | ~42 tok/s ModelFit est. |
| Llama 3.1 405B Instruct405B | Q4_K_M | 250 GB | ~4.1 tok/s slow ModelFit est. |
| Llama 4 Maverick400B MoE | Q4_K_M | 252 GB | ~76 tok/s ModelFit est. |
| Ornith 1.5 122B125.09B | Q4_K_M | 79 GB | ~107 tok/s ModelFit est. |
| Ling-3.0-flash124B | Q4_K_M | 81 GB | ~152 tok/s ModelFit est. |
| Devstral 2 123B123B | Q4_K_M | 78 GB | ~13 tok/s ModelFit est. |
| Qwen3.5 122B-A10B Instruct122B MoE | Q4_K_M | 75 GB | ~102 tok/s ModelFit est. |
| Laguna S 2.1118B | Q4_K_M | 99 GB | ~85 tok/s ModelFit est. |
| Llama 4 Scout109B MoE | Q4_K_M | 70 GB | ~73 tok/s ModelFit est. |
| Qwen3-Next 80B-A3B80B MoE | Q8_0 | 88 GB | ~201 tok/s ModelFit est. |
| Kolibri 178.1B | Q4_K_M | 52 GB | ~181 tok/s ModelFit est. |
| Llama 3.1 70B Instruct70B | Q4_K_M | 44 GB | ~22 tok/s ModelFit est. |
| DeepSeek-R1 Distill Llama 70B70B | Q4_K_M | 44 GB | ~22 tok/s ModelFit est. |
| Mixtral 8x7B Instruct46.7B MoE | Q4_K_M | 32 GB | ~70 tok/s ModelFit est. |
| Flint 40B-A2.5B39.8B MoE | NVFP4 | 24 GB | ~236 tok/s ModelFit est. |
| Sage-A1.535.95B | Q4_K_M | 23 GB | ~225 tok/s ModelFit est. |
| Qwen3.5 35B-A3B Instruct35B MoE | Q8_0 | 41 GB | ~160 tok/s ModelFit est. |
| Ornith 1.0 35B35B | Q4_K_M | 23 GB | ~228 tok/s ModelFit est. |
| Nanbeige4.3 33B33.1B | Q8_0 | 37 GB | ~27 tok/s ModelFit est. |
| Poolside Laguna XS.233B | Q4_K_M | 25 GB | no estimate |
| Laguna XS 2.133B | Q4_K_M | 22 GB | ~215 tok/s ModelFit est. |
| Qwen3 32B32B | Q8_0 | 37 GB | ~27 tok/s ModelFit est. |
| DeepSeek-R1 Distill Qwen 32B32B | Q8_0 | 37 GB | ~27 tok/s ModelFit est. |
| Olmo 3.1 32B Instruct32B | Q4_K_M | 21 GB | ~45 tok/s ModelFit est. |
| Qwen3 30B30B | Q8_0 | 32 GB | ~130 tok/s ModelFit est. |
| NVIDIA Nemotron Cascade 2 30B-A3B30B MoE | Q6_K | 26 GB | no estimate |
| Cohere North Mini Code30B | Q4_K_M | 20 GB | no estimate |
| Nemotron 3.5 Lightning 30B-A3B30B MoE | Q4_K_M | 27 GB | ~160 tok/s ModelFit est. |
| GLM-4.7-Flash 30B-A3B30B MoE | Q4_K_M | 20 GB | ~175 tok/s ModelFit est. |
| Granite Switch 4.1 30B30B | Q4_K_M | 23 GB | ~42 tok/s ModelFit est. |
| Muse Glimmer 30B29.8B | Q4_K_M | 18.3 GB | ~56 tok/s ModelFit est. |
| Granite 4.2 30B29.3B | Q4_K_M | 17.9 GB | ~49 tok/s ModelFit est. |
| AREX-2 27B27.4B | Q4_K_M | 18.6 GB | ~54 tok/s ModelFit est. |
| Gemma 2 27B Instruct27B | Q4_K_M | 23 GB | ~50 tok/s ModelFit est. |
| Gemma 3 27B Instruct27B | Q4_K_M | 23 GB | ~52 tok/s ModelFit est. |
| Qwen3.5 27B Instruct27B | Q8_0 | 29 GB | ~34 tok/s ModelFit est. |
| Qwen3.6 27B27B | Q8_0 | 32 GB | ~34 tok/s ModelFit est. |
| Jebadiah 27B27B | Q8_0 | 31 GB | ~34 tok/s ModelFit est. |
| Mistral Small 3.124B | Q8_0 | 25 GB | ~36 tok/s ModelFit est. |
| LFM2 24B-A2B Instruct24B MoE | Q4_K_M | 15.3 GB | ~239 tok/s ModelFit est. |
| Devstral Small 2 24B24B | Q4_K_M | 16.6 GB | ~58 tok/s ModelFit est. |
| Mistral Small 22B22B | Q4_K_M | 18.4 GB | ~65 tok/s ModelFit est. |
| Qwen2.5 14B Instruct14B | Q4_K_M | 12.3 GB | ~89 tok/s ModelFit est. |
| Phi-3 Medium 14B14B | Q4_K_M | 12.3 GB | no estimate |
| Phi-4 14B14B | Q8_0 | 15.9 GB | ~56 tok/s ModelFit est. |
| Qwen2.5 Coder 14B14B | Q4_K_M | 12.3 GB | ~89 tok/s ModelFit est. |
| Qwen3 14B14B | Q8_0 | 17.3 GB | ~57 tok/s ModelFit est. |
| DeepSeek-R1 Distill Qwen 14B14B | Q8_0 | 16.0 GB | ~57 tok/s ModelFit est. |
| Ministral 3 14B13.5B | Q4_K_M | 10.3 GB | ~89 tok/s ModelFit est. |
| Mellum2.1 12B-A2.5B Thinking12.1B MoE | Q4_K_M | 9.3 GB | no estimate |
| Mistral Nemo 12B12B | Q8_0 | 13.4 GB | ~67 tok/s ModelFit est. |
| Gemma 3 12B Instruct12B | Q4_K_M | 10.7 GB | ~102 tok/s ModelFit est. |
| Gemma 4 12B12B | Q8_0 | 14.1 GB | ~66 tok/s ModelFit est. |
| GLM-4-9B-Chat9.4B | Q8_0 | 11.2 GB | ~88 tok/s ModelFit est. |
| Gemma 2 9B Instruct9B | Q4_K_M | 8.2 GB | ~109 tok/s ModelFit est. |
| Ornith 1.0 9B9B | Q4_K_M | 6.7 GB | ~143 tok/s ModelFit est. |
| MiMo-V2.6-Distill-Qwen-9B9B | Q4_K_M | 6.9 GB | ~139 tok/s ModelFit est. |
| Jebadiah 9B9B | Q8_0 | 11.0 GB | ~94 tok/s ModelFit est. |
| Granite 4.2 8B8.8B | Q4_K_M | 6.1 GB | ~130 tok/s ModelFit est. |
| Granite Switch 4.1 8B8.8B | Q4_K_M | 7.6 GB | ~110 tok/s ModelFit est. |
| MiniCPM-V 4.5 8B8.7B | Q4_K_M | 6.8 GB | ~139 tok/s ModelFit est. |
| Ministral 3 8B8.4B | Q4_K_M | 7.1 GB | ~123 tok/s ModelFit est. |
| LFM2.5 8B-A1B8.3B MoE | Q4_K_M | 6.6 GB | ~277 tok/s ModelFit est. |
| Llama 3.1 8B Instruct8B | Q8_0 | 9.2 GB | ~96 tok/s ModelFit est. |
| Qwen3 8B8B | Q8_0 | 9.3 GB | ~93 tok/s ModelFit est. |
| Granite 4.1 8B Instruct8B | Q4_K_M | 6.6 GB | ~126 tok/s ModelFit est. |
| Mistral 7B Instruct7B | Q8_0 | 8.4 GB | ~100 tok/s ModelFit est. |
| Qwen2.5 Coder 7B7B | Q4_K_M | 6.6 GB | ~153 tok/s ModelFit est. |
| DeepSeek-R1 Distill Qwen 7B7B | Q8_0 | 8.7 GB | ~103 tok/s ModelFit est. |
| Gemma 4 E4B4.5B | Q8_0 | 8.7 GB | ~143 tok/s ModelFit est. |
| Qwen3.5 4B Instruct4B | Q8_0 | 5.4 GB | ~153 tok/s ModelFit est. |
| NVIDIA Nemotron 3.5 Bolt E4B4B | Q4_K_M | 9.1 GB | ~171 tok/s ModelFit est. |
| Phi-3 Mini 3.8B3.8B | Q4_K_M | 4.3 GB | no estimate |
| Phi-4 Mini 3.8B3.8B | Q8_0 | 4.9 GB | ~152 tok/s ModelFit est. |
| Granite 4.2 3B3.7B | Q4_K_M | 3.2 GB | ~228 tok/s ModelFit est. |
| Ministral 3 3B3.4B | Q4_K_M | 4.1 GB | ~187 tok/s ModelFit est. |
| Qwen2.5 3B Instruct3B | Q4_K_M | 3.6 GB | ~248 tok/s ModelFit est. |
| Llama 3.2 3B Instruct3B | Q4_K_M | 3.6 GB | ~225 tok/s ModelFit est. |
| Granite 4.1 3B Instruct3B | Q4_K_M | 3.0 GB | ~228 tok/s ModelFit est. |
| Gemma 4 E2B2.3B | Q8_0 | 5.7 GB | ~248 tok/s ModelFit est. |
| Gemma 2 2B Instruct2B | Q4_K_M | 2.8 GB | ~243 tok/s ModelFit est. |
| Qwen3.5 2B Instruct2B | Q4_K_M | 2.8 GB | ~249 tok/s ModelFit est. |
| LFM3 1.7B1.73B | Q4_K_M | 2.1 GB | ~322 tok/s ModelFit est. |
| Qwen2.5 1.5B Instruct1.5B | Q4_K_M | 2.5 GB | ~327 tok/s ModelFit est. |
| Llama 3.2 1B Instruct1B | Q4_K_M | 2.0 GB | ~345 tok/s ModelFit est. |
| Gemma 3 1B Instruct1B | Q8_0 | 2.0 GB | ~328 tok/s ModelFit est. |
| Qwen3.5 0.8B Instruct0.8B | Q4_K_M | 1.8 GB | ~355 tok/s ModelFit est. |
| Qwen2.5 0.5B Instruct0.5B | Q4_K_M | 1.8 GB | ~415 tok/s ModelFit est. |
| SmolLM2 360M0.36B | Q4_K_M | 1.5 GB | ~350 tok/s ModelFit est. |
Speeds come from ModelFit's estimates (CC BY 4.0) where it covers the model, otherwise from our bandwidth formula. Method
Too big for this machine
DeepSeek-R1 671B (390 GB).
Questions
What is the best AI model for the M5 Ultra (512 GB, 80-core GPU)?
Qwen3.8-Flash-Next at Q4_K_XL: the largest model on our shortlist of current models that runs at 4-bit or better at a usable speed (about 111 tok/s, ModelFit est.).
How much memory can a model use on the M5 Ultra (512 GB, 80-core GPU)?
About 384 GB of the 512 GB.